This is the classic case, and I have and have seen many such needs. For text-based stuff as you describe, it's fairly straightforward. One of the groty issues we've run into with all available solutions, however, is that images (logos, etc.) from the web page get re-interpolated and this just makes them look like the dog barfed. Since there are very few pages any more without some graphics, this becomes a more pressing issue. I think converting is more important than manual formatting, but maybe I'm biased because we run into this all the time.

As to XML versus CSS versus a little language, I think that if it "just works" for the primary use case, any extra that I/we have to learn in any format is not too much to ask.